4.11
Using Symmetric MultiProcessing (SMP) on a Massively Parallel Processing (MPP) system to Support Meso-Scale Weather Modeling for the Global Theater Weather Analysis and Prediction System (GTWAPS)
Keith H. North, IBM, Omaha, NE; and J. D. Benson, J. Tuccillo, and R. T. Williams
The utilization of Symmetric MultiProcessing (SMP) on a Massively Parallel Processing (MPP) system along with other tools represent the computing backbone of the GTWAPS at AFWA. During the past two years, GTWAPS has moved from a two frame SP (single system) environment to 18 frames (comprising 5 systems). These systems are networked together using a High Performance Gateway Node (HPGN) which allows very high speed data transfer between nodes. System reliability is enhanced through the use of GPFS running on a RAID-5 disk subsystem. AFWA is currently running 18 theaters per day, generating over 80GB of output data which is then transformed into over 90,000 GIF images. These images are made available to meteorologists in the field through AFWIN (Air Force Weather Information Network) over the Internet. A very high degree of success has been obtained by parallelizing GTWAPS applications and resulting in significant decreases in application run times.
Session 4, Application of IIPS in Forecasting (Parallel with Session 3)
Monday, 10 January 2000, 1:00 PM-5:45 PM
Previous paper Next paper